skshmjn/CRAG-EVAL

CRAG-EVAL是一个用于评估文档相关性的数据集，通过二分类方式确定给定查询或上下文中检索到的文档是否相关。数据集由问题和文档的对组成，每个对都被标注为相关或不相关，并附有由模型BAAI/bge-large-en生成的相似度分数。该数据集可用于训练或评估二分类相关性分类器，微调用于语义相关性的Transformer模型，以及文档重排系统的基准测试。

CRAG-EVAL is a dataset for evaluating document relevance through binary classification, determining whether a retrieved document is relevant or not in the context of a given query. The dataset consists of pairs of questions and documents, each labeled as relevant or not relevant, and accompanied by a similarity score generated by the model BAAI/bge-large-en. It is useful for training or evaluating binary relevance classifiers, fine-tuning transformer models for semantic relevance, and benchmarking document reranking systems.